publications

2026

1. ICLR

   LLM-as-a-Prophet: Understanding Predictive Intelligence with Prophet Arena

   Qingchuan (Tony) Yang^*, Simon Mahns^*, Sida Li^*, Anri Gu^*, Jibang Wu, and Haifeng Xu

   International Conference on Learning Representations (ICLR), 2026

   Featured in The Atlantic, Yahoo Finance, The Dispatch, and Thinking Machines Lab’s Inkling release · 450k+ views on X.

   Abstract arXiv BibTeX Website

   Prophet Arena is a live prediction benchmark that evaluates large language models on their ability to retrieve and aggregate evidence for probabilistic forecasting.

   @article{yang2026prophet,
     title = {LLM-as-a-Prophet: Understanding Predictive Intelligence with Prophet Arena},
     author = {Yang, Qingchuan (Tony) and Mahns, Simon and Li, Sida and Gu, Anri and Wu, Jibang and Xu, Haifeng},
     journal = {International Conference on Learning Representations (ICLR)},
     year = {2026},
     note = {<span style="color: #cc0000;">Featured in <a href="https://www.theatlantic.com/technology/2026/02/ai-prediction-human-forecasters/685955/" style="color: #cc0000; text-decoration: underline;">The Atlantic</a>, <a href="https://finance.yahoo.com/news/ai-now-matches-prediction-markets-230346352.html" style="color: #cc0000; text-decoration: underline;">Yahoo Finance</a>, <a href="https://thedispatch.com/article/ai-forecasting-national-security-intelligence/" style="color: #cc0000; text-decoration: underline;">The Dispatch</a>, and <a href="https://thinkingmachines.ai/news/introducing-inkling/" style="color: #cc0000; text-decoration: underline;">Thinking Machines Lab's Inkling release</a> &middot; <a href="https://x.com/ProphetArena/status/1956928877106004430" style="color: #cc0000; text-decoration: underline;">450k+ views on X</a>.</span>}
   }

2. ICML

   EPSVec: Efficient and Private Synthetic Data Generation via Data Vectors

   Amin Banayeeanzade^*, Qingchuan (Tony) Yang^*, Deqing Fu, Spencer Hong, Erin Babinsky, Alfy Samuel, Anoop Kumar, Robin Jia, and Sai Praneeth Karimireddy

   International Conference on Machine Learning (ICML), 2026

   arXiv BibTeX

   @article{banayeeanzade2026epsvec,
     title = {EPSVec: Efficient and Private Synthetic Data Generation via Data Vectors},
     author = {Banayeeanzade, Amin and Yang, Qingchuan (Tony) and Fu, Deqing and Hong, Spencer and Babinsky, Erin and Samuel, Alfy and Kumar, Anoop and Jia, Robin and Karimireddy, Sai Praneeth},
     journal = {International Conference on Machine Learning (ICML)},
     year = {2026},
   }

3. Sampling More, Getting Less: Calibration is the Diversity Bottleneck in LLMs

   Amin Banayeeanzade^*, Qingchuan (Tony) Yang^*, Dhruv Tarsadiya, Fatemeh Bahrani, Leonardo Blas, Robin Jia, Meisam Razaviyayn, and Sai Praneeth Karimireddy

   In submission, 2026

   arXiv BibTeX

   @article{banayeeanzade2026sampling,
     title = {Sampling More, Getting Less: Calibration is the Diversity Bottleneck in LLMs},
     author = {Banayeeanzade, Amin and Yang, Qingchuan (Tony) and Tarsadiya, Dhruv and Bahrani, Fatemeh and Blas, Leonardo and Jia, Robin and Razaviyayn, Meisam and Karimireddy, Sai Praneeth},
     journal = {In submission},
     year = {2026},
   }

4. Psychological Steering via Calibrated Activation Injections

   Leonardo Blas, Qingchuan (Tony) Yang, Robin Jia, and Emilio Ferrara

   In submission, 2026

   BibTeX

   @article{blas2026steering,
     title = {Psychological Steering via Calibrated Activation Injections},
     author = {Blas, Leonardo and Yang, Qingchuan (Tony) and Jia, Robin and Ferrara, Emilio},
     journal = {In submission},
     year = {2026}
   }

^* Equal contribution.